If the VM is stuck in the destroying state, we create pages for both the
runner and the VM. Since the VM is already being destroyed, there’s no
need to raise the runner destroy deadline. I will help with the page
storms when the host is down.
The destroying state also includes the prevent_destroy state, so a
single check is sufficient.
We generally wait 10m for destroy pages, so I decreased it for runners
from 15m to 10m as well.
5f1df7e45d